👾 We are hiring a NodeJS Systems Engineer in New York. Apply here.
LM Studio Logo

LM Studio Beta Releases

🪲 🗣️ Please report bugs and feedback to:

Beta and Experimental Releases

VersionBuildOSArchLast UpdatedDownload URL
0.3.26macOSarm642024-09-25LM-Studio-0.3.2-arm64.dmg (apple silicon)
0.3.26Windowsx86_642024-09-25LM-Studio-0.3.2-Setup.exe (x86)
0.3.26Linuxx86_642024-09-25LM_Studio-0.3.2.AppImage (x86)
0.3.26Windowsarm642024-09-25LM-Studio-0.3.2-Setup.exe (arm64)
LM Studio is provided under the terms of use.

Release Notes - LM Studio 0.3.2 Beta

Build 6

  • Presets are BACK!
    • Legacy presets are automatically migrated to the new format upon save. The old files are NOT deleted.
    • Load parameters are not currently included in the new preset format. Favor editing the model's default config in My Models.
  • Show live token count for user input using the selected model's tokenizer
  • Show live token count for system prompt using the selected model's tokenizer
  • New buttons at the top of the chat: Clear All Messages, Clone Chat
  • Open App Settings using [⌘,] or [Ctrl + ,] from anywhere
    • Settings icon is now in the bottom right corner of the app, no longer in the sidebar
    • [⌘4] or [Ctrl + 4] now opens My Models
  • LM Studio now has a Beta in-app update channel, in addition to the Stable channel
    • Choose between Stable and Beta in the App Settings
    • Beta versions might have the same version number as Stable versions, but the build number will be higher
  • ThemeSelector ([⌘K ⌘T] or [Ctrl + K, Ctrl + T] from anywhere)
  • Stats underneath input box: show context fullness % (click on the label to toggle to n_tokens/n_context)
  • QoL: Chat auto-name, skip if time to first token high or tok/sec low
  • Perf: switching between long chats should be faster
  • First model download suggestion: suggest Gemma 2B for machines with less than 12GB of VRAM
  • Copy debug info to clipboard by right clicking the gear icon next to the model

Bug Fixes

  • Bug fix: Fix GGUF parsing bugs for Qwen 2.5, Llama 3.2, c4ai and more models
  • Bug fix: [Windows] fix app close/minimize/maximize buttons background overflow
  • Bug fix: Helpful error when the first message is longer than the total context size
  • Bug fix: Button to show / hide intermediate steps in RAG sometimes did not work (fixed)
  • Bug fix: [Developer] Checkbox for showing debug info did not work when unchecked (fixed)
  • Bug fix: Open downloads folder will open the custom downloads folder instead of the default one
  • Bug fix: Updated Jinja template parser to support models using more complex templates
  • Bug fix: Esc wouldn't close some dialogs